The multichannel mechanism in color vision was experimentally studied. Psychophysical measurements were carried out using DOG(Difference of Gaussians) test patterns of the combination of dual colors such as red-green and blue-yellow. Difference in numbers and spatial characteristics of channel mechanisms were observed when the colors in the center and the surrounding of the test patterns are reversed.

View full abstract

Observation of stereoscopic images in 3D-display give to eyes fatigue. The primary factor is that the visual functions of eyes have an ambalance status by stereoscopic observation. Namely , eyes fatigue give an ambalance status to accommodation and in this study , we were examined on convergence by extreme stereoscopic effect. As a result stereoscopic account of eye soft is grasp of stereoscopic vision. image was observed to no eye fatigue points is select of kind with distance.

View full abstract

For a European non-kanji Japanese language learner who uses the alphabet to write his mother language, especially a beginner, kanji character cognition is an important problem. Accordingly we studied, mainly through eye movement, the effect of kanji structure instruction on kanji character cognition from the initial period. The eye movement after learning shows that the fixations gather in the center of the parts that roughly construct one kanji and, for a new kanji, there is a tendency for the fixations to be concentrated in the unknown parts. Exchanges of known parts also occurred, as the results of a writing test show, which indicates the learner memorizes one kanji as a union of parts. Therefore, kanji structure instruction is effective in increasing the ability to recognize kanji character.

View full abstract

Sequentially displaying patterns containing specific key-details, we asked subjects to link and order them by a rule that they must define, and we analysed their eye-movements during encoding learning and recognition tasks. The global evolution of the eye movements starts from the discovery of all the visible information and is then more and more concentrated on the most informative parts. However, "concentration and informative parts" should be defined in relation to the nature of the rule, the subject's ability and the quality of visual analysis. But overall, the ability to analyse the patterns using peripheral vision is the most important variable. Different classes of eye movements could be observed: a "global" class (center of the pattern), a "local" class (key-detail and center) and a "total" class (all the details). The relative performances are very different, creating a scale [ global > local > total ], showing the great importance of peripheral vision in this task.

View full abstract

On the observation of TV screen, not only images projected to foveal retina but also ones to peripheral retina are seen. Studies on peripheral color vision are reviewed. Although there exist differences between central and peripheral color vision, recent studies say that foveal like color vision can be obtained by increasing stimulus size with eccentricity, or under light adaptation. It corresponds to that the relative numbers of L and M cones remains constant over wide range of eccentricities.

View full abstract

An objective picture quality scale for television pictures is studied. The following impairment factors are investigated. (1) Weighted noise by three dimensional human visual characteristics (2) Weighted noise by human visual masking functions (3) Block distortions in the digital coding pictures using DCT As a result, the objective picture quality scale and the assessment quality of the subjective pictures obtained experimentally show good agreement.

View full abstract

This paper presents a new method of face identification using a two dimensional physical model named Potential Net and Karhunen-Loeve (K-L) expansion. Using Potential Net, invariant features can be extracted from input images under a given transformation group such as rotation by which the categories associated with the input images are kept invariant. It consists of nodes in two dimensional grid configuration and the configuration is deformed by image forces which move the nodes to the position of facial organs. As the shape of the deformed net includes facial features, their features are efficient for face identification. A high dimensional feature space of invariant features obtained by Potential Net is translated into an effective low dimensional feature space of discriminant features by adopting K-L expansion. We have estimated the stability of our method for faces with facial expressions, different illumination conditions and input data with head variations arid show the effectiveness of the proposed method.

View full abstract

Image retrieval system providing an effective access to desired data in image database has been studied. Since image is more ambiguous than language, an image retrieval that uses only keyword as retrieval key is not enough. In this paper, we propose an image retrieval method that uses a rough sketch as retrieval key. Image in a database system is divided into segments and is described by color, position, size, shape, and geographical relation of each area. These structural features are represented by attributed relational graph. Similarity is measured by graph matching between their graphs and a graph derived from rough sketch.

View full abstract

Kalman filter design used for the restoration of motion-blurred images assumes the knowledge of signal statics and the blurring system. This paper discusses the estimation of parameters required in Kalman filter design for the restoration of motion-blurred images. It is shown that the innovation of the Kalman filter can be used to improve the estimations.

View full abstract